Xpert Magic: Automated Data Pipeline

View on GitHub | View App

Role: Epidemiologist, Data Lead

Position: Postdoctoral Fellowship @ University of Manitoba

Domain: Health Informatics, Clinical Research

1 The Challenge

In the SWOP HPV Study in Nairobi, GeneXpert machines provided critical diagnostic data for HPV and CTNG. However, the raw output was semi-structured and difficult to aggregate, creating a bottleneck for longitudinal epidemiological analysis. Manual data cleaning was unsustainable for a high-volume clinical setting.

2 The Solution

I developed Xpert Magic, a production-grade Shiny application that automates the transformation of GeneXpert raw outputs into tidy, analysis-ready datasets.

• Automated Parsing: Built custom R logic to extract diagnostic results from heterogeneous machine files.
• Data Integrity: Implemented real-time validation checks to ensure consistency between HPV and CTNG results.
• User-Centric Design: Designed a simple “Upload and Export” interface tailored for lab staff without R experience.

3 Technical Deep-Dive

• Modular R Code: Organized using a clean directory structure for scalability and maintenance.
• Tech Stack: R, Shiny, tidyverse, readxl, DT, bs4Dash, route.
• Impact: Eliminated manual data entry errors and reduced the clinical data-to-analysis lead time by over 90%.